Professional Documents
Culture Documents
J Ast 2006 10 011 PDF
J Ast 2006 10 011 PDF
www.elsevier.com/locate/aescte
Abstract
The determination of drag and lift of transport aircraft from wind tunnel measurements is discussed for compressible as well as incompressible
flow. Effects of model support are neglected. No simplifying assumptions on the flow along the test section walls are made. Based on the
assumption of correctable wind tunnel flow, balance measurements and wake survey measurements together indicate a way to derive an angle of
attack correction towards the in-flight condition. This will require the same level of (sufficient) accuracy for both types of measurements and for
the propulsion mass flux. The blockage correction is assumed to be known. Buoyancy correction is briefly discussed. Propulsion is fully taken
into account.
Wake survey computational postprocessing is treated in detail. It is shown that wake survey measurements can be limited to the wake cross-
sectional area.
© 2007 Elsevier Masson SAS. All rights reserved.
Keywords: Transport aircraft; Wind tunnel measurements; Wake survey measurements; Drag and lift analysis; Propulsion effect; Transonic flow
Nomenclature
which constitutes the firm belief that the near-field flow about Iv+w = − ρuū dS (3)
the model aircraft is in itself correct, but corresponds to a dif-
SD
ferent angle of attack and a different freestream velocity if the
model aircraft is considered to be “in-flight”.
Ii = − ρu(u − u∞ − ū) + (p − p∞ ) dS (4)
The present analysis indicates that balance measurements
SD
and wake survey measurements together are in principle a suf-
ficient basis to derive the angle of attack correction from the
J =− ρuw dS (5)
assumption of correctable flow, and also to say more about the
drag/thrust diagnostics than was previously possible. SD
ū = ū − u∞ (6)
2. Balance equations: aircraft in flight
with ū given by the requirement
Based on the Reynolds-averaged Navier–Stokes equations,
the balance equations in the flight-direction and in the lift- p[ū, v = 0, w = 0] = p∞ (7)
direction are respectively, compare [9],
and ṁ is the mass flux produced by the propulsion units, as
XB = D − Tx = Iv+w + Ii − ṁu∞ (1) caused by fuel consumption or a T.P.S. in the wind tunnel.
Here the downstream control surface SD is chosen normal to
ZB = L − Tz = J (2)
the flight direction and sufficiently far downstream that viscous
where stresses can effectively be neglected. Compare Fig. 1.
J. van der Vooren / Aerospace Science and Technology 12 (2008) 337–345 339
In the absence of propulsion (Tx = 0, ṁ = 0), Eq. (1) re- In the special case of incompressible flow, there holds
duces to
2
XB = D = Iv+w + Ii (8) ū = (pt − pt∞ ) + u∞ 2 (14)
ρ∞
Iv+w is then the sum of viscous drag and wave drag, while Ii is
the induced drag [9]. where pt is the total pressure,
In the present analysis it is convenient to introduce small ve- 1 1
locity disturbances with respect to the parallel flow u = ū, v = p − p∞ = −ρ∞ ū∗ u − ρ∞ ∗ u2 − ρ∞ v 2 + w 2 (15)
0, w = 0, p = p∞ , i.e. 2 2
ρ = ρ̄ = ρ∞ (16)
∗ u = u − ū = ū0[ε]; v, w = ū0[ε] (9)
Note that Eq. (15) matches the compressible expression Eq. (11)
with ε being small. The value of ū can be obtained from Eq. (7) for M = 0, ρ̄ = ρ∞ .
and is
Upon using Eqs. (11), (6), (9), the expression for Ii , Eq. (4),
2H 2 (γ −1)/γ can be written as
ū = u∞ 1 + − es/R − 1 (10)
u∞ 2 (γ − 1)M∞ 2
1 2 1
Ii = ρ̄ v + w 2 − ρ̄ 1 − M 2 ∗ u2 dS + 0 ε 3
Note that the value of ū depends on the increments H and 2 2
SD
s in respectively stagnation enthalpy and entropy relative to
(17)
their values in the undisturbed flow. In case of propulsion, nei-
ther of these increments can be assumed to be small, and this is If v = 0, w = 0, there is no streamline curvature in SD , such
the reason for the particular small disturbance assumptions in that p = p∞ and consequently ∗ u = 0, see Eq. (11). This in-
Eq. (9). dicates that 12 ρ̄(v 2 + w 2 ) is the leading term in Eq. (11) and
The result for compressible flow is thus logically
1 1
p − p∞ = −ρ̄ ū∗ u − ρ̄ 1 − M 2 ∗ u2 − ρ̄ v 2 + w 2 ū∗ u = 0 v 2 + w 2 = ū2 0 ε 2 (18)
2 2
in retrospect, compare Eq. (9).
+ 0 ε3 (11)
∗u It follows, that
ρ = ρ̄ − ρ̄M 2 + 0 ε2 (12) 1 2
ū Ii = ρ̄ v + w 2 dS (19)
where 2
SD
ρ̄ = ρ[ū, v = 0, w = 0, H, s]; is an accurate approximation of Ii for compressible flow, that
M = M[ū, v = 0, w = 0, H, s] (13) will be used throughout in the further analysis.
340 J. van der Vooren / Aerospace Science and Technology 12 (2008) 337–345
In the special case of incompressible flow, Eq. (19) reduces extended flow and the real flow over the boundary layer thick-
to ness in the downstream control plane (SD ), as well as the effect
1 2 of the difference between uniform onset flow (u∞ ) and the real
Ii = ρ∞ v + w 2 dS (20) flow in the entire upstream control plane (SU ). Note also that
2
SD their values are independent of the location of SU for a fixed lo-
compare Eq. (16). cation of SD , implying fixed values of I˜v+w , I˜i , J, while also
ṁu∞ and XB , ZB are fixed. Compare Eqs. (21), (22).
3. Balance equations: aircraft model in the wind tunnel Explicit expressions for Dtunnel , Ltunnel are not given, be-
cause they will appear not to be relevant in the further analysis.
In this case the balance equations in the flight-direction (di- Just like for the aircraft in flight, Eq. (24) will be approxi-
rection of the undisturbed flow in the test section) and in the mated in the further analysis by
Along the same lines as with Eq. (30), the expression (31) through SP 1 + SP 2 (the viscous/thermodynamic losses pro-
can be written as duced by the jet engines (T.P.S.) or the propellers). Note that
B cos α + X
B sin α these last losses must be measurable in the wind tunnel with
J =Z (35)
acceptable accuracy in order to obtain Dv+w .
Here X B , Z
B are known from the balance measurements, ṁ is
known from the T.P.S. setting, u∞ can be obtained from the 7. Induced drag in flight
wind tunnel setting, and Iv+w , Ii , J can be calculated from
wake survey measurements as will be described in the further Again following [9] the induced drag is defined by
analysis. Note that Eqs. (34), (35) are precisely the integral
1
forms of momentum conservation in, respectively normal to, Di = I i + ρ v 2 + w 2 nx dS (41)
the flight direction, applied to the aircraft (model) in flight, and 2
SP 1 +SP 2
quantified using wind tunnel balance and wake survey measure-
ments together. compare Eqs. (19), (20). Di is thus basically the difference
In principle, either of Eqs. (34), (35) can be used to calcu- between the crossflow kinetic energy in SD (produced by the
late α. However, Eq. (35) is not recommended because it is propelled aircraft) and the production of crossflow kinetic en-
ill-conditioned for calculating α (J ≈ Z B , cos α ≈ 1). The ergy on SP 1 + SP 2 (produced by the jet engines (T.P.S.) or the
best way to calculate α is most likely to solve Eqs. (34), propellers). Note the similarity with Eq. (40). The assumption
(35) simultaneously in closed form. Upon dividing Eq. (34) by of correctable wind tunnel flow then implies that the flow in
Eq. (35), it can be shown that α can be calculated from SP 1 + SP 2 is correctly simulated in
SP 1 +
SP 2 in the wind tun-
B nel. And since also Ii can be obtained from wake survey mea-
X
tg(α + β) = (36) surements as indicated before, Di can, in principle, be obtained
B
Z correctly from wind tunnel flow data. However, this requires
where the auxiliary angle β is defined by that the necessary flow data in SP 1 + SP 2 can be measured in
Iv+w + Ii − ṁu∞ the wind tunnel with acceptable accuracy. Note also that these
tg β = (37) data must be adapted to the angle of attack correction α.
J
It is also instructive to observe that, since α 1, β 1, 8. Buoyancy correction
Eqs. (36), (37) imply with second order accuracy
B
X Iv+w + Ii − ṁu∞ In case of buoyancy in the empty test section of the wind
α = − (38)
b
Z J tunnel, it is generally assumed that there exists a small, mea-
surable pressure gradient in the axial (x−) direction. Naturally,
Two special cases can finally be mentioned. For X B = 0, the
there are also small deviations from the uniaxial flow field that
model aircraft is in equilibrium in the flight direction “in the have some influence. However, in practice, it is not possible to
wind tunnel”, see Eq. (21). Then α = −β. For Iv+w + Ii − quantify this influence. Hence, it is common practice to correct
ṁu∞ = 0, see Eq. (1), the same is true “in flight”, giving β = 0, only for the axial pressure gradient, but even then in an approx-
B /Z
tg α = X b .
imate simple manner.
It is not unusual to limit buoyancy correction to correcting
6. Viscous drag and wave drag in flight the pressure drag of the fuselage, being by far the largest body
of the aircraft configuration. In this case, the total drag of the
Following [9] the sum of viscous drag and wave drag is de- model aircraft in the wind tunnel is diminished by D, calcu-
fined by lated from some simple formula like e.g.
Dv+w = − div(ρūq ) dV (39) =f p
D Vfus. (42)
x
Vv+w
where Vfus. is the volume of the aircraft model, p/x is the
In view of the assumption of correctable wind tunnel flow, the
v+w = Vv+w is correct and hence Dv+w can, in prin- (average) pressure gradient over the fuselage length and f is a
flow in V
shape factor.
ciple, be obtained correctly from wind tunnel flow data.
The following procedure suffices to effectuate the buoy-
Upon combining Eqs. (28), (39) there can be written by
ancy correction in the preceding analysis. Simply replace D
D − D and consequently XB by XB − D, compare Eq. (21),
Dv+w = Iv+w − ρū(q · n) dS (40)
before calculating α.
SP 1 +
SP 2
This shows that Dv+w is the difference between the trans- 9. Wake survey analysis
port of −ū through the downstream control plane SD (the
viscous/thermodynamic losses produced by the propelled air- Wake survey analysis requires the numerical calculation of
craft model), compare Eqs. (3), (23), and the transport of −ū the following integrals, compare Eqs. (3), (19), (5),
J. van der Vooren / Aerospace Science and Technology 12 (2008) 337–345 343
Iv+w = − ρuū dS (43) Next consider σ . It follows from mass conservation that
∂u 1 ∂ρ ∂ρ ∂ρ
SD σ =− − u +v +w (53)
1 2 ∂x ρ ∂x ∂y ∂z
Ii = ρ̄ v + w 2 dS (44)
2 Upon using Eqs. (6), (9), (12), (18), this can be approximated
SD by
J =− ρuw dS (45) ∂ū
σ =− + 0 ε2 (54)
SD
∂x
using also that ρ̄ is constant along streamlines. Because ū = 0
where ρ = ρ̄ = ρ∞ in the special case of incompressible flow. outside the wake, Eq. (54) shows that σ is indeed effectively
Outside the (inviscid) wake there holds (by definition) zero outside the wake, while inside the wake the approximation
H = 0, s = 0 (pt = pt∞ in the special case of incompress-
∂ū
ible flow), such that there ū = u∞ and hence ū = 0, compare σ =− (55)
Eqs. (10), (14), (6). And since the flow is irrotational outside ∂x
the wake, there holds also ω
= 0. suffices.
Consequently, Eq. (43) reduces to The right-hand side of Eq. (50) can be written as
∂(ρ̄v) ∂(ρ̄w) ∂ ρ̄ ∂ ρ̄
Iv+w = − ρuū dS (46) − = −ρ̄ωx + v −w (56)
∂z ∂y ∂z ∂y
Swake and similarly the right-hand side of Eq. (51) takes the form
where Swake ⊂ SD is the intersection of the wake and SD . Each ∂(ρ̄v) ∂(ρ̄w) ∂ ρ̄ ∂ ρ̄
quantity in Eq. (46) can be obtained from wake survey mea- + = ρ̄σ + v +w (57)
∂y ∂z ∂y ∂z
surements and hence Iv+w can be calculated numerically. Note
Since ρ̄ = ρ∞ outside the wake, compare Eq. (13), while ωx
again that I˜v+w = Iv+w , though its value is not required in the
and σ are (effectively) zero, Eqs. (56), (57) show that also the
present analysis.
right-hand sides of Eqs. (50), (51) are (effectively) zero outside
Consider the expression for Ii , Eq. (44). Upon assuming
the wake.
∂ϕ ∂ψ ∂ϕ ∂ψ It follows from the above discussion that Eq. (48) reduces to
ρ̄v = + , ρ̄w = − (47)
∂y ∂z ∂z ∂y 1 1
Ii = (ψωx − ϕσ ) dS − (ψqt − ϕqn ) d∂S (58)
the expression for Ii can be written as 2 2
Swake ∂SD
1 1
Ii = (ψωx − ϕσ ) dS − (ψqt − ϕqn ) d∂S (48) where according to e.g. [4]
2 2
SD ∂SD 1
ψ[y, z] = f [η, ζ ] ln r dη dζ (59)
where ∂SD is the outer boundary of SD , 2π
Swake
∂w ∂v ∂v ∂w with
ωx = − , σ= + (49)
∂y ∂z ∂y ∂z ∂(ρ̄v) ∂(ρ̄w)
f [η, ζ ] = − (60)
ψ and ϕ must be solved from ∂z ∂y
∂ 2ψ ∂ 2ψ ∂(ρ̄v) ∂(ρ̄w) and similarly
+ 2 = − (50)
∂y 2 ∂z ∂z ∂y 1
ϕ[y, z] = g[η, ζ ] ln r dη dζ (61)
2π
∂ 2 ϕ ∂ 2 ϕ ∂(ρ̄v) ∂(ρ̄w)
+ 2 = + (51) Swake
∂y 2 ∂z ∂y ∂z with
and ∂(ρ̄v) ∂(ρ̄w)
g[η, ζ ] = + (62)
qt = wny − vnz , qn = vny + wnz (52) ∂y ∂z
r being given by
are respectively the tangential velocity and the normal veloc-
ity on ∂SD , with n pointing outwards and t being right-turning r = (y − η)2 + (z − ζ )2 (63)
with respect to the x-axis.
Entirely in the spirit of wake survey measurements, which Since ∂SD is at infinity, Ii is subsequently calculated from
are indeed commonly restricted to the wake cross-section, it 1
Ii = (ψωx − ϕσ ) dS (64)
will be shown that ωx and σ , as well as the right-hand sides of 2
Eqs. (50), (51), are (effectively) zero outside the wake cross- Swake
section Swake . upon assuming that the boundary integral in Eq. (58) vanishes.
As to ωx this is evident from the fact that the flow is irrota- However, this is true only if the following two conditions are
tional outside the wake. satisfied,
344 J. van der Vooren / Aerospace Science and Technology 12 (2008) 337–345
∂(ρ̄v) ∂(ρ̄w)
− dy dz = −ρ∞ (wny − vnz ) d∂S 10. Discussion and conclusions
∂z ∂y
Swake ∂Swake
The assumption of correctable flow is at the basis of model
= −ρ∞ ωx dS = 0 (65)
testing of transport aircraft in the wind tunnel. Provided that the
Swake model is small enough compared with the test section dimen-
and similarly sions, correctable flow allows the translation of measured data
in the wind tunnel to flight conditions by correcting the angle
∂(ρ̄v) ∂(ρ̄w)
+ dy dz = ρ∞ (vny + wnz ) d∂S of attack and the flight speed (blockage correction). Until not
∂y ∂z
Swake ∂Swake so long ago, only balance measurements were performed to ob-
tain the aerodynamic forces acting on the model aircraft. In case
= ρ∞ σ dS = 0 (66) of propulsion, this implied that lift and drag could be obtained
Swake only if the corresponding thrust components were known. Com-
If these conditions are not fulfilled, the boundary integral in pare Eqs. (21), (22). At that time, drag diagnostics could not be
Eq. (58) will be infinite, in disagreement with the physical re- given and corrections on angle of attack and flight speed were
quirement of finite crossflow kinetic energy. It is therefore ac- often a matter of concern.
ceptable to say that these conditions are indeed satisfied. They A major step forward was made with the upcoming of
imply respectively that the total amount of trailing vorticity, (two- and three-dimensional) wake survey measurements. This
generated by the aircraft and its propulsion, is zero, while the enabled to discriminate between viscous/thermodynamic and
wake also generates no mass in the transverse plane SD . crossflow kinetic energy losses in a downstream measuring
Note that the calculation of I˜i is not discussed, because its plane perpendicular to the undisturbed flow in the empty wind
value is not required in the present analysis. However, it can tunnel. These losses are the prime components of drag minus
be calculated using proper boundary conditions for ψ and ϕ on thrust as measured in the wind tunnel, though other components
∂SD . may exist. Compare Eq. (21).
Calculation of J requires the introduction of the auxiliary The present analysis of wind tunnel measurements of trans-
potential port aircraft is valid for compressible as well as incompressible
φ=y (67) flow and makes no simplifying assumptions on the test section
flow as in [1–3,7,8]. Also there is no restriction on the mag-
and the auxiliary vector nitude of propulsion induced flow perturbations H , s as
A = [Ay , Az ]T , Ay = ρuw, Az = −ρuv (68) in [7]. Test section buoyancy is briefly discussed.
The analysis indicates that balance measurements and wake
Then Eq. (45) can be written as
survey measurements together are a sufficient basis to derive the
∂φ ∂φ angle of attack correction required in translating wind tunnel
J =− Ay + Az dS (69)
∂y ∂z conditions to in-flight conditions, provided that balance data,
SD
wake survey data, and also the propulsion mass flux, have the
which can be worked out to obtain same level of (sufficient) accuracy. Wake survey computational
∂ρw ∂ρv ∂u ∂u post-processing is treated in detail. It is shown that all integrals
J= yu − dS + yρ w −v dS
∂y ∂z ∂y ∂z over the measuring plane can (with good approximation) be re-
SD SD stricted to the wake cross-sectional area, such that the wake
− yρuqt d∂S (70) survey can be limited to that area.
Following [9], the determination of viscous drag plus wave
∂SD
drag and of induced drag in flight has been briefly discussed.
To leading order, this expression can be approximated by, com- Either of these physical drag components is basically the differ-
pare Eqs. (12), (18), ence between certain losses produced by the propelled aircraft
∂ ρ̄w ∂ ρ̄v ∂ ū ∂ ū and the corresponding losses produced by the propulsion alone.
J=∼ yu − dS + yρ w −v dS
∂y ∂z ∂y ∂z Compare Eqs. (40), (41). Only the losses produced by the pro-
Swake Swake pelled aircraft, viz. the viscous/thermodynamic losses and the
(71) crossflow kinetic energy losses in the measuring plane, can be
since the boundary integral in Eq. (70) can be shown to vanish obtained from wake survey measurements. However, it does not
in view of Eqs. (65), (66), and ρ̄ = ρ∞ , ū = u∞ , ωx = 0 outside seem easy to measure the corresponding losses produced by the
Swake . propulsion alone, such as by a throughflow nacelle, a T.P.S., or
Note that the calculation of J is not discussed, because a simulated propeller. Yet it is important to know these losses.
its value is not required in the present analysis. However, it Otherwise, viscous drag plus wave drag and induced drag can-
can be approximated by specifying ρqt = (ρqt )e ∼ = (ρ̄qt )e = not be obtained and this will hamper the translation of wind
∂ψ
ρ∞ (qt )e = ∂ϕ
∂t − ∂n and using the boundary conditions for ψ tunnel data to true flight where real jet engines and propellers
and ϕ on ∂ SD . are being used.
J. van der Vooren / Aerospace Science and Technology 12 (2008) 337–345 345
Acknowledgements [4] C.P. van Dam, K. Nikfetrat, Accurate prediction of drag using Euler meth-
ods, J. Aircraft 29 (3) (1992) 516–519, Engineering Notes.
[5] D. Destarac, Far-field/near-field drag balance and applications of drag ex-
The author wants to express his thanks to Daniel Destarac traction in CFD, in: VKI Lecture Series 2003, CFD-Based Aircraft Drag
(ONERA) and Jan van Egmond (NLR, retired) for the help, in- Prediction and Reduction, National Institute of Aerospace, Hampton, VA,
terest and encouragement during the preparation of this paper November 3–7, 2003.
in its present form. [6] E.B. Klunker, Contribution to methods for calculating the flow about thin
lifting wings at transonic speeds – Analytic expressions for the far field,
NASA TN D-6530, November 1971.
References [7] K. Kusunose, Advanced wake integration method for experimental drag
prediction, in: VKI Lecture Series 2002/2003, CFD-Based Aircraft Drag
[1] S. Amant, Drag prediction and decomposition from wake surveys and cal- Prediction and Reduction, February 3–7, 2003.
culations in subsonic flows, AIAA 2001–2446. [8] E.C. Maskell, Progress towards a method for the measurement of the com-
[2] A. Betz, A method for the direct determination of profile drag, ponents of the drag of a wing of finite span, R.A.E. Technical Report 72232,
Z.F.M. 16 (42) (1925) (in German). 1972.
[3] G.W. Brune, Quantitative three-dimensional low-speed wake surveys, in: [9] J. van der Vooren, D. Destarac, Drag/thrust analysis of jet-propelled tran-
Fifth Symposium on Numerical and Physical Aspects of Aerodynamic sonic transport aircraft; definition of physical drag components, Aerospace
Flows, California State University, Long Beach, CA, 13–15 January 1992. Science & Technology 8 (6) (2004) 545–556.